Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userEmploying low-pass filtered temporal speech features for the training of ideal ratio mask in speech enhancement
توظيف ميزات الكلام الزمنية المرحلة المنخفضة لتدريب قناع نسبة مثالية في تحسين الكلام
296
0 0
0.0
(
0
)
Created by
Shamra Editor
Ask ChatGPT about the research
تتابع طريقة تحسين الكلام القائم على اخفاء قناع مضاعف ينطبق على الطيفية من الكلام الفاسد من ضوضاء الإدخال، وغالبا ما تستخدم شبكة عميقة العصبية (DNN) لتعلم القناع. على وجه الخصوص، يمكن أن تكون الميزات الشائعة الاستخدام للتعرف على الكلام التلقائي بمثابة مدخلات DNN لتعلم القناع حسن التصرف الذي يقلل بشكل كبير من تشويه الضوضاء للكلمات المعالجة. تقترح هذه الدراسة إعادة معالجة ميزات خطاب المدخلات لمقنعة النسبة المثالية (IRM) - DNN بواسطة Lowpass Filtering من أجل تخفيف مكونات الضوضاء. على وجه الخصوص، فإننا نوظف تحويل المويجات المنفصلة (DWT) لتحلل تسلسل ميزة الكلام الزمنية وتوسيع نطاق معاملات التفاصيل، مما يتوافق مع الجزء المرتفع من التسلسل. تكشف التجارب الأولية التي أجراها مجموعة فرعية من تيميت كوربوس أن الطريقة المقترحة يمكن أن تجعل IRM الناتجة تحقيق جودة أعلى للكلام وضوحا للإشارات الاضافة عن الضوضاء الخليفة مقارنة مع IRM الأصلي، مما يشير إلى أن تسلسل الميزات الزمنية المرشح ل Lowpass يمكن أن يتعلم متفوقة شبكة IRM لتعزيز الكلام.
The masking-based speech enhancement method pursues a multiplicative mask that applies to the spectrogram of input noise-corrupted utterance, and a deep neural network (DNN) is often used to learn the mask. In particular, the features commonly used for automatic speech recognition can serve as the input of the DNN to learn the well-behaved mask that significantly reduce the noise distortion of processed utterances. This study proposes to preprocess the input speech features for the ideal ratio mask (IRM)-based DNN by lowpass filtering in order to alleviate the noise components. In particular, we employ the discrete wavelet transform (DWT) to decompose the temporal speech feature sequence and scale down the detail coefficients, which correspond to the high-pass portion of the sequence. Preliminary experiments conducted on a subset of TIMIT corpus reveal that the proposed method can make the resulting IRM achieve higher speech quality and intelligibility for the babble noise-corrupted signals compared with the original IRM, indicating that the lowpass filtered temporal feature sequence can learn a superior IRM network for speech enhancement.
References used
https://aclanthology.org/
rate research
Read More
In this project we study wavelet and wavelet transform, and the possibility of its employment in the processing and analysis of the speech signal in order to enhance the signal and remove noise of it. We will present different algorithms that depend
on the wavelet transform and the mechanism to apply them in order to get rid of noise in the speech, and compare the results of the application of these algorithms with some traditional algorithms that are used to enhance the speech.
In this paper, we describe experiments designed to evaluate the impact of stylometric and emotion-based features on hate speech detection: the task of classifying textual content into hate or non-hate speech classes. Our experiments are conducted for
three languages -- English, Slovene, and Dutch -- both in in-domain and cross-domain setups, and aim to investigate hate speech using features that model two linguistic phenomena: the writing style of hateful social media content operationalized as function word usage on the one hand, and emotion expression in hateful messages on the other hand. The results of experiments with features that model different combinations of these phenomena support our hypothesis that stylometric and emotion-based features are robust indicators of hate speech. Their contribution remains persistent with respect to domain and language variation. We show that the combination of features that model the targeted phenomena outperforms words and character n-gram features under cross-domain conditions, and provides a significant boost to deep learning models, which currently obtain the best results, when combined with them in an ensemble.
The speech recognition is one of the most modern technologies, which entered force
in various fields of life, whether medical or security or industrial techniques. Accordingly,
many related systems were developed, which differ from each otherin fea
ture extraction
methods and classification methods.
In this research,three systems have been created for speech recognition.They differ
from each other in the used methods during the stage of features extraction.While the first
system used MFCC algorithm, the second system used LPCC algorithm, and the third
system used PLP algorithm.All these three systems used HMM as classifier.
At the first, the performance of the speechrecognitionprocesswas studied and
evaluatedfor all the proposedsystems separately. After that, the combination algorithm was
applied separately on eachpair of the studied system algorithmsin order to study the effect
of using the combination algorithm onthe improvement of the speech recognition process.
Twokinds of errors(simultaneous errors and dependent errors) were usedto evaluate
the complementaryof each pair of the studied systems, and to study the effectiveness of the
combination on improving the performance of speech recognition process. It can be seen
from the results of the comparison that the best improvement ratio of speech recognition
has been obtained in the case of collection MFCC and PLP algorithms with recognition
ratio of 93.4%.
Bias mitigation approaches reduce models' dependence on sensitive features of data, such as social group tokens (SGTs), resulting in equal predictions across the sensitive features. In hate speech detection, however, equalizing model predictions may
ignore important differences among targeted social groups, as hate speech can contain stereotypical language specific to each SGT. Here, to take the specific language about each SGT into account, we rely on counterfactual fairness and equalize predictions among counterfactuals, generated by changing the SGTs. Our method evaluates the similarity in sentence likelihoods (via pre-trained language models) among counterfactuals, to treat SGTs equally only within interchangeable contexts. By applying logit pairing to equalize outcomes on the restricted set of counterfactuals for each instance, we improve fairness metrics while preserving model performance on hate speech detection.
This paper describes the submission to the IWSLT 2021 Low-Resource Speech Translation Shared Task by IMS team. We utilize state-of-the-art models combined with several data augmentation, multi-task and transfer learning approaches for the automatic s
peech recognition (ASR) and machine translation (MT) steps of our cascaded system. Moreover, we also explore the feasibility of a full end-to-end speech translation (ST) model in the case of very constrained amount of ground truth labeled data. Our best system achieves the best performance among all submitted systems for Congolese Swahili to English and French with BLEU scores 7.7 and 13.7 respectively, and the second best result for Coastal Swahili to English with BLEU score 14.9.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
